The High-Performance Task Runner for Python Backends. Parallel. Deterministic. Fast.
Zetten is a dependency-aware execution engine designed to unify how you run tests, linters, and builds. It ensures that your workflow remains identical across local development environments and any CI platform, only faster.
Modern Python projects often require coordinating various tools (tests, type-checkers, formatters). Zetten eliminates "Glue Code Fatigue" by providing:
- Parallel Execution: Automatically identifies independent tasks and runs them concurrently across your CPU cores.
- Three-Tier Variable System: Advanced command templating with a strict priority hierarchy: CLI Flags (-k) > Config Vars > Environment Variables.
- Parallel Execution: Automatically identifies independent tasks and runs them concurrently across your CPU cores.
- Three-Tier Variable System: Advanced command templating with a strict priority hierarchy: CLI Flags (-k) > Config Vars > Environment Variables.
- Smart Caching: Uses content-addressable hashing to skip tasks if their specific inputs haven't changed since the last run.
- Platform Agnostic: Behaves identically on macOS, Windows, Linux, or any CI/CD provider.
- Dependency Awareness: Define a Directed Acyclic Graph (DAG) of tasks to ensure correct execution order (e.g.,
setupalways precedestest).
Zetten is designed to be the "API" for ANY AI agent. It comes with built-in configuration files for:
- Cursor (
.cursorrules) - Windsurf (
.windsurfrules) - GitHub Copilot (
.github/copilot-instructions.md) - Claude Code (
CLAUDE.md)
These files automatically point your AI to docs/ai_context.md, ensuring it knows exactly how to build, test, and run your project without hallucinating commands.
"Zetten allows me to just tell Cursor 'run the tests' and it knows exactly what to do, efficiently."
- ⚡ Worker Pool Concurrency: Maximizes resource usage by running non-dependent tasks in parallel.
- 🏷️ CI Tagging: Execute logical groups of tasks (e.g.,
run --tag ci) with a single command. - 🛡️ Failure Propagation: If a foundational task fails, Zetten halts downstream execution to prevent cascading errors.
- 🔍 Intelligent Diagnostics: Includes
ztn doctorto identify environment inconsistencies instantly. - ⏱️ Performance Analytics: (Coming Soon) Real-time insights into time saved via parallelism.
Zetten is built for speed. Benchmarks against popular task runners show it provides the fastest developer experience for incremental builds.
| Metric | Tool | Time | vs Zetten |
|---|---|---|---|
| Startup | ztn |
2.08 ms | 1.00x |
| (CLI overhead) | just |
2.15 ms | 1.04x |
make |
3.85 ms | 1.85x | |
poe |
41.88 ms | 20.15x | |
| Smart Caching | ztn |
3.49 ms | 1.00x |
| (No-op re-run) | just |
4.26 ms | 1.22x |
make |
5.63 ms | 1.61x | |
poe |
64.68 ms | 18.52x |
> Benchmarks run on macOS (Apple Silicon). Startup measures tool --version. Smart Caching measures time to detect no input changes and skip execution.
See Full Methodology & Detailed Results
Install Zetten:
pip install zettenInitiate a project:
ztn initFor Python script execution:
[tool.zetten.tasks.hello]
script = "my_module:main" # Runs: python -c "import my_module; my_module.main()"
description = "Run a python function"Define tasks in pyproject.toml:
[tool.zetten.tasks.lint]
cmd = "ruff check src"
inputs = ["src/"]
tags = ["ci"]
[tool.zetten.tasks.test]
cmd = "pytest"
depends_on = ["lint"]
inputs = ["src/", "tests/"]
tags = ["ci"]
[tool.zetten.tasks.build]
description = "Build the project"
# Supports Fallback Syntax: ${VAR:-default}
cmd = "mkdir -p ${build_dir} && python -m build --outdir ${DEST:-dist}"
depends_on = ["lint"]
inputs = ["src/"]Define tasks in zetten.toml:
[tasks.setup]
cmd = "pip install -r requirements.txt"
[tasks.lint]
cmd = "ruff check src"
inputs = ["src/"]
tags = ["ci"]
[tasks.test]
cmd = "pytest"
depends_on = ["setup"]
inputs = ["src/", "tests/"]
tags = ["ci"]Run tasks:
ztn run test
ztn run lint testZetten will only re-run tasks when their inputs change.
Zetten uses a deterministic three-tier system to resolve variables:
- Tier 1 (CLI):
ztn run build -k build_dir=output(Highest Priority) - Tier 2 (Config): Values defined in
[tool.zetten.vars] - Tier 3 (Env): System environment variables (e.g.,
$USER,$PATH)
Zetten is designed for the modern CI/CD pipeline. By using Tags and Strict Mode, you can ensure your pipeline is both flexible and safe.
# Force a specific version and environment in CI
ztn run --tag ci -k VERSION=${GITHUB_SHA} -k ENV=prodIf a foundational task fails, Zetten halts downstream execution immediately to save CI minutes and prevent cascading failures.
Configuration is explicit by design:
- No templating
- No conditionals
- No implicit behavior
Configuration lives in:
pyproject.toml(preferred)zetten.toml(for legacy or minimal projects)
If no configuration is found, Zetten will explain how to resolve the issue.
ztn run <tasks>— Execute tasks with parallel dependency resolution.ztn run <task> -k KEY=VAL— Override any variable via the CLI.ztn watch <tasks>— Precision re-runs on input changes.ztn graph— Visualizes the Directed Acyclic Graph (DAG) of your tasks.ztn doctor— Diagnoses configuration and environmental health issues.ztn init— Interactive project setup and template generation.
Zetten is currently in v1.3.2. If no configuration file is found, Zetten will provide clear instructions on how to initialize your project.
Full documentation is available at: docs.zetten.in
We love Rust and Python! If you want to help make Zetten even faster:
- Fork the repo.
- Add your feature (don't forget the tests!).
- Open a Pull Request.
Built with ❤️ for the Python community using the speed of Rust.
Please open an issue or discussion on GitHub before proposing large changes.